Robamine

Python code and OpenAI Gym environments for testing RL algorithms with models of ARL's robots.

Under heavy development

Here you can find the change log for all the releases. We use the Semantic Versioning Specifications for release numbering.

The documentation for the latest release will be found in the future here. For now see at the end of this README on how to generate the docs locally.

Installation

Mujoco

In order to work with mujoco_py you should have Mujoco version 1.50 with its license in ~/.mujoco such that the structure of ~/.mujoco would be:

.mujoco
 |-- mjkey.txt
 |-- mjpro150
     |--bin
     |-- doc
     |-- include
     |-- model
     |-- sample

Install a Python Virtual Env

Install glew if not installed:

sudo apt install libglew-dev

Install python3-tk:

sudo apt-get install python3-tk

sudo pip install virtualenv
virtualenv ~/robamine-env --python=python3 --prompt='[robamine env] '

At the end of the ~/robamine-env/bin/activate script, add the following lines:

If you have Nvidia:

Check with nvidia-smi which nvidia driver is installed (e.g. 384) and replace 384 with your nvidia driver in the lines below:

# If the virtualenv inherits the `$PYTHONPATH` from your system:
export PYTHONPATH="$VIRTUAL_ENV/lib"

# For using Mujoco:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/nvidia-384
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.mujoco/mjpro150/bin
export LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libGLEW.so:/usr/lib/nvidia-384/libGL.so

If you do not have Nvidia:

# If the virtualenv inherits the `$PYTHONPATH` from your system:
export PYTHONPATH="$VIRTUAL_ENV/lib"

# For using Mujoco:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.mujoco/mjpro150/bin
export LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libGLEW.so:/usr/lib/x86_64-linux-gnu/libGL.so

Then, activate the environment (you have to activate it each time you want to use it):

source ~/robamine-env/bin/activate

Install Robamine

Activate your virtual environment and then clone the repository and install the package:

git clone https://github.com/iSaran/robamine.git
cd robamine
pip install -e .

Using ROS environments

In case you want to use the ROS environments (e.g. LWRROS-v1) with ROS kinetic you should install the following packages in order for ROS to work with Python 3:

sudo apt-get install python3-pip python3-yaml
sudo pip3 install rospkg catkin_pkg

Remove from your activate script of your environment this line:

export PYTHONPATH="$VIRTUAL_ENV/lib"

And finally, uncomment the commented imports of the ROS environments from robamine/envs/__init__.py.

Run

To run with GUI simply run the following script (recommended):

python run.py

To run without GUI:

python run.py --no-gui --yml=YAML_FILE

Run unit tests

After major changes run every unit test:

python run_tests.py

Generate documentation

Useful information can be found in the documentation which can be generated locally by:

cd doc
make html

It will be build in doc/_build/index.html.

Development

Follow the steps below in order to add a new algorithm or environment:

Add the Python class of the agent/environment in algo or env.
Add its name in yaml/available.yml, its default values as a yaml file in yaml/defaults and value constraints in yaml/constraints.

iSaran / robamine